The question of whether the mantle convects as a single layer, or whether the convection is
split vertically into two or more layers, is perhaps the most fundamental unanswered
question concerning the structure and dynamics of the mantle, and has ramifications
throughout solid Earth geosciences. The most likely mechanism for enforcing layered
convection is the action of the major phase transitions of the transition zone, and in
particular, the strong endothermic phase transition between 
-spinel and
perovskite+magnesiowustite that occurs at around 660 km depth. Because this transition is
endothermic, it is deflected downwards in cold downwellings, and upwards in hot
upwellings, resulting in a mass anomaly that impedes the penetration of the downwelling or
upwelling. In the past few years, there has been a huge resurgence of interest and modeling
activity devoted to studying the influence of the 400 and 660 km phase transitions, and the
effects identified in these studies may provide the key to reconciling various seemingly
contradictory geophysical observations.
The theoretical framework to studies of mantle convection with phase transitions was laid down over two decades ago [ Schubert and Turcotte, 1971; Schubert et al., 1975], and extended by the analytical models of plumes encountering an endothermic phase boundary presented by Olson and Yuen [1982]. Peltier et al. [1989] extended the earlier stability analyses to more realistic spherical geometry and updated parameters. However, at the high Rayleigh number (an indicator of convective vigor) characteristic of the Earth, the effects are too complex to be treated with analytical models, and time-dependent numerical simulations are necessary.
Christensen and Yuen [1984, 1985] presented the first and most definitive numerical
models of phase-transition modulated mantle convection, performed in a two-dimensional
box, and the major findings in these works have stood up in all subsequent studies. These
findings are (1) the anomalous buoyancy due to deflection of the phase transition in an
up- or down-welling is much more important than the effect of latent heat release, and thus the
net effect of an endothermic phase transition is to oppose the flow of material across itself,
(2) the propensity to layering increases with increasing Rayleigh number and increasing
negativity of the Clapeyron slope (the gradient of the phase boundary in
pressure/temperature space), (3) in the regime between complete layering and completely
whole-mantle convection, intermittent layering can occur [ Christensen and Yuen,
1985], (4) variable viscosity in the downwelling slab does not appear to have a large effect
on the strength of phase change necessary to stop it from penetrating [ Christensen and
Yuen, 1984, 1985]. However, after these pioneering studies there was no further work
on the subject for many years, due partly to the unrealistically large value of the Clapeyron
slope these authors found was necessary to enforce complete layering (-6 MPa K
),
and partly to the continuing uncertainty as to whether the 660 km discontinuity represented
a phase or compositional discontinuity.